Extending Functionality: mdisp

Shawn Pearce <>
Fri Mar 21 05:34:17 CET 2003


Chris Pressey <> wrote:
> If my hunch *is* on target, then mdisp might serve as a good basis for a
> compromise between processes and objects, such as what might be desirable
> for a GUI, or a simulated world.  Any thoughts on this?


One - why do we have to put processes asleep?

What's fundamentally wrong with erts that is causing us to consider
paging state/context on and off of processes?  This is smelling a
whole lot like what we would have to do in Java to accomplish the
same task, only we can currently create more processes (threads)
before the emulator faults, and there is the fancy thing called
ETS.  :)

What I'm getting at is, why can't these be normal processes and let
erts handle the very concept of them not being alive, or executing
on the sender's call stack, etc.  And then can we fix erts?

I've been watching this thread with interest, wondering what will come
of it.  I'm praying its not another bastard semi-threaded /
semi-not-threaded system. :)

-- 
Shawn.

  Test-tube babies shouldn't throw stones.



More information about the erlang-questions mailing list